The Pulse
Only 2% of Agent-Platform Users Picked for Model Alignment
In VentureBeat Intelligence’s August survey, 3 of 162 respondents using a listed agent orchestration platform said alignment with a leading AI model most influenced their choice. Flexibility across models and tools, reliability, ease of dev
AI.info Team · Written with AI ·
Just 3 of 162 enterprise respondents who use a listed agent orchestration platform said the underlying AI model was the main reason they chose one, according to VentureBeat Intelligence’s August survey. The report, published October 5, found that model alignment, which it calls “model gravity,” drew 2% of responses; performance, including latency and memory, ranked lower at 1%. Flexibility across models and tools led at 22%, followed by production reliability at 20%, ease of development at 19% and control over agent execution at 18%.
“Only 2% of platform users named ‘model gravity.’”
VentureBeat Intelligence’s authors write.
Model ownership does not settle platform choice
OpenAI, Anthropic and Google all offer platforms for building and running agents alongside their models. Yet the survey suggests enterprises weigh what those platforms let teams do more heavily than whether a platform comes bundled with a preferred model. VentureBeat calls the model-alignment option “model gravity” and defines it as choosing a platform for its native fit with a leading model, such as Anthropic’s tooling for Claude.
The 2% figure comes from the 162 respondents who said they use a listed orchestration platform, not from all 166 qualified survey respondents. Respondents chose one leading factor. Security and permissions drew 10%, total cost of ownership 9%, and performance—including latency and memory—1%. The report’s rounded shares for flexibility, reliability, development ease and execution control add to 79%, while its summary describes those four factors as 78% of choices.
Platforms are used without always becoming primary
Provider-built platforms still appear widely in the survey’s usage answers. OpenAI’s Agents SDK or Responses API was used by 45% of all respondents, Google’s Enterprise Agent Platform by 39%, and Anthropic’s Claude Platform by 27%. Those figures allow respondents to name multiple platforms, so they describe reported use, not exclusive market share.
Use also does not always make a provider’s platform the enterprise’s primary choice. Among respondents who use Anthropic’s Claude Platform, 38% named it primary; for OpenAI’s platform, the figure was 69%. VentureBeat notes a limitation in the Anthropic comparison: the survey named Claude Platform and Agent Skills in its usage question, but Claude Agent SDK and Managed Agents in its question about platforms under consideration. The report says that when it excludes respondents who named a primary platform they had not marked as used, the Claude figure rises to 41%.
Many buyers expect another platform decision
Sixty percent of all 166 respondents plan to adopt a new, additional or replacement orchestration platform within 12 months, and 28% expect to do so within three months. Among enterprises already running two or more platforms, 84% plan a change, compared with 46% of those running one. Those plans point to active procurement, but they do not establish that firms are abandoning provider platforms or switching for any single reason.
For enterprises planning a change, OpenAI was the most frequently named platform under consideration, at 40%; Anthropic’s Claude Agent SDK or Managed Agents followed at 35%, and Google’s platform at 29%. Respondents could select several candidates, and 11% said they were evaluating without a shortlist. The report also found that 67% expect their agent control plane to sit at least partly outside a provider-managed service by the end of 2026, including those favoring hybrid, external or in-house arrangements.
A survey snapshot, not a market census
VentureBeat Intelligence received 221 responses in August; 169 passed its qualifying questions, and it removed three whose answers contradicted one another. The remaining 166 respondents came from organizations with at least 100 employees. Half identified as final decision-makers for AI purchases, while 32% said they were recommenders or influencers.
The report describes its audience as self-selected VentureBeat readers, not a probability sample, so the results should be read as a snapshot of that group rather than a market-wide estimate. Month-to-month comparisons also involve different respondents: the report says industry, company size and purchasing roles differed between its July and August samples. Within those limits, the clearest procurement signal is that just three of 162 platform users put the model behind their platform first.